; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
; RUN: opt -passes='lower-matrix-intrinsics' -data-layout='p:128:128' -S < %s | FileCheck %s --check-prefix=PTR128
; RUN: opt -passes='lower-matrix-intrinsics' -data-layout='p:64:64' -S < %s | FileCheck %s --check-prefix=PTR64
; RUN: opt -passes='lower-matrix-intrinsics' -data-layout='p:32:32' -S < %s | FileCheck %s --check-prefix=PTR32

; To properly support the matrix intrinsics on, e.g., 32-bit platforms (without
; the need to emit `libc` calls), we perform strided index calculations using
; the same pointer bit-width as the matrix pointers, as determined by the data
; layout. To verify this behaviour, this test runs several strided loads and
; stores through the lowering pass with (32|64|128)-bit pointers, and verifies
; the generated code extends / truncates strides accordingly. Similarly,
; `data-layout-multiply-fused.ll` adopts this approach to verify the same
; behaviour for index calculations emitted while lowering fused matrix
; multiplies.

define <9 x double> @strided_load_3x3_i128(ptr %in, i128 %stride) {
; PTR128-LABEL: @strided_load_3x3_i128(
; PTR128-NEXT:  entry:
; PTR128-NEXT:    [[VEC_START:%.*]] = mul i128 0, [[STRIDE:%.*]]
; PTR128-NEXT:    [[VEC_GEP:%.*]] = getelementptr double, ptr [[IN:%.*]], i128 [[VEC_START]]
; PTR128-NEXT:    [[COL_LOAD:%.*]] = load <3 x double>, ptr [[VEC_GEP]], align 8
; PTR128-NEXT:    [[VEC_START1:%.*]] = mul i128 1, [[STRIDE]]
; PTR128-NEXT:    [[VEC_GEP2:%.*]] = getelementptr double, ptr [[IN]], i128 [[VEC_START1]]
; PTR128-NEXT:    [[COL_LOAD3:%.*]] = load <3 x double>, ptr [[VEC_GEP2]], align 8
; PTR128-NEXT:    [[VEC_START4:%.*]] = mul i128 2, [[STRIDE]]
; PTR128-NEXT:    [[VEC_GEP5:%.*]] = getelementptr double, ptr [[IN]], i128 [[VEC_START4]]
; PTR128-NEXT:    [[COL_LOAD6:%.*]] = load <3 x double>, ptr [[VEC_GEP5]], align 8
; PTR128-NEXT:    [[TMP0:%.*]] = shufflevector <3 x double> [[COL_LOAD]], <3 x double> [[COL_LOAD3]], <6 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5>
; PTR128-NEXT:    [[TMP1:%.*]] = shufflevector <3 x double> [[COL_LOAD6]], <3 x double> poison, <6 x i32> <i32 0, i32 1, i32 2, i32 poison, i32 poison, i32 poison>
; PTR128-NEXT:    [[TMP2:%.*]] = shufflevector <6 x double> [[TMP0]], <6 x double> [[TMP1]], <9 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8>
; PTR128-NEXT:    ret <9 x double> [[TMP2]]
;
; PTR64-LABEL: @strided_load_3x3_i128(
; PTR64-NEXT:  entry:
; PTR64-NEXT:    [[STRIDE_CAST:%.*]] = trunc i128 [[STRIDE:%.*]] to i64
; PTR64-NEXT:    [[VEC_START:%.*]] = mul i64 0, [[STRIDE_CAST]]
; PTR64-NEXT:    [[VEC_GEP:%.*]] = getelementptr double, ptr [[IN:%.*]], i64 [[VEC_START]]
; PTR64-NEXT:    [[COL_LOAD:%.*]] = load <3 x double>, ptr [[VEC_GEP]], align 8
; PTR64-NEXT:    [[VEC_START1:%.*]] = mul i64 1, [[STRIDE_CAST]]
; PTR64-NEXT:    [[VEC_GEP2:%.*]] = getelementptr double, ptr [[IN]], i64 [[VEC_START1]]
; PTR64-NEXT:    [[COL_LOAD3:%.*]] = load <3 x double>, ptr [[VEC_GEP2]], align 8
; PTR64-NEXT:    [[VEC_START4:%.*]] = mul i64 2, [[STRIDE_CAST]]
; PTR64-NEXT:    [[VEC_GEP5:%.*]] = getelementptr double, ptr [[IN]], i64 [[VEC_START4]]
; PTR64-NEXT:    [[COL_LOAD6:%.*]] = load <3 x double>, ptr [[VEC_GEP5]], align 8
; PTR64-NEXT:    [[TMP0:%.*]] = shufflevector <3 x double> [[COL_LOAD]], <3 x double> [[COL_LOAD3]], <6 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5>
; PTR64-NEXT:    [[TMP1:%.*]] = shufflevector <3 x double> [[COL_LOAD6]], <3 x double> poison, <6 x i32> <i32 0, i32 1, i32 2, i32 poison, i32 poison, i32 poison>
; PTR64-NEXT:    [[TMP2:%.*]] = shufflevector <6 x double> [[TMP0]], <6 x double> [[TMP1]], <9 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8>
; PTR64-NEXT:    ret <9 x double> [[TMP2]]
;
; PTR32-LABEL: @strided_load_3x3_i128(
; PTR32-NEXT:  entry:
; PTR32-NEXT:    [[STRIDE_CAST:%.*]] = trunc i128 [[STRIDE:%.*]] to i32
; PTR32-NEXT:    [[VEC_START:%.*]] = mul i32 0, [[STRIDE_CAST]]
; PTR32-NEXT:    [[VEC_GEP:%.*]] = getelementptr double, ptr [[IN:%.*]], i32 [[VEC_START]]
; PTR32-NEXT:    [[COL_LOAD:%.*]] = load <3 x double>, ptr [[VEC_GEP]], align 8
; PTR32-NEXT:    [[VEC_START1:%.*]] = mul i32 1, [[STRIDE_CAST]]
; PTR32-NEXT:    [[VEC_GEP2:%.*]] = getelementptr double, ptr [[IN]], i32 [[VEC_START1]]
; PTR32-NEXT:    [[COL_LOAD3:%.*]] = load <3 x double>, ptr [[VEC_GEP2]], align 8
; PTR32-NEXT:    [[VEC_START4:%.*]] = mul i32 2, [[STRIDE_CAST]]
; PTR32-NEXT:    [[VEC_GEP5:%.*]] = getelementptr double, ptr [[IN]], i32 [[VEC_START4]]
; PTR32-NEXT:    [[COL_LOAD6:%.*]] = load <3 x double>, ptr [[VEC_GEP5]], align 8
; PTR32-NEXT:    [[TMP0:%.*]] = shufflevector <3 x double> [[COL_LOAD]], <3 x double> [[COL_LOAD3]], <6 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5>
; PTR32-NEXT:    [[TMP1:%.*]] = shufflevector <3 x double> [[COL_LOAD6]], <3 x double> poison, <6 x i32> <i32 0, i32 1, i32 2, i32 poison, i32 poison, i32 poison>
; PTR32-NEXT:    [[TMP2:%.*]] = shufflevector <6 x double> [[TMP0]], <6 x double> [[TMP1]], <9 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8>
; PTR32-NEXT:    ret <9 x double> [[TMP2]]
;
entry:
  %load = call <9 x double> @llvm.matrix.column.major.load.v9f64.i128(ptr %in, i128 %stride, i1 false, i32 3, i32 3)
  ret <9 x double> %load
}

define <9 x double> @strided_load_3x3_const_stride_i128(ptr %in) {
; PTR128-LABEL: @strided_load_3x3_const_stride_i128(
; PTR128-NEXT:  entry:
; PTR128-NEXT:    [[COL_LOAD:%.*]] = load <3 x double>, ptr [[IN:%.*]], align 8
; PTR128-NEXT:    [[VEC_GEP:%.*]] = getelementptr double, ptr [[IN]], i128 16
; PTR128-NEXT:    [[COL_LOAD1:%.*]] = load <3 x double>, ptr [[VEC_GEP]], align 8
; PTR128-NEXT:    [[VEC_GEP2:%.*]] = getelementptr double, ptr [[IN]], i128 32
; PTR128-NEXT:    [[COL_LOAD3:%.*]] = load <3 x double>, ptr [[VEC_GEP2]], align 8
; PTR128-NEXT:    [[TMP0:%.*]] = shufflevector <3 x double> [[COL_LOAD]], <3 x double> [[COL_LOAD1]], <6 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5>
; PTR128-NEXT:    [[TMP1:%.*]] = shufflevector <3 x double> [[COL_LOAD3]], <3 x double> poison, <6 x i32> <i32 0, i32 1, i32 2, i32 poison, i32 poison, i32 poison>
; PTR128-NEXT:    [[TMP2:%.*]] = shufflevector <6 x double> [[TMP0]], <6 x double> [[TMP1]], <9 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8>
; PTR128-NEXT:    ret <9 x double> [[TMP2]]
;
; PTR64-LABEL: @strided_load_3x3_const_stride_i128(
; PTR64-NEXT:  entry:
; PTR64-NEXT:    [[COL_LOAD:%.*]] = load <3 x double>, ptr [[IN:%.*]], align 8
; PTR64-NEXT:    [[VEC_GEP:%.*]] = getelementptr double, ptr [[IN]], i64 16
; PTR64-NEXT:    [[COL_LOAD1:%.*]] = load <3 x double>, ptr [[VEC_GEP]], align 8
; PTR64-NEXT:    [[VEC_GEP2:%.*]] = getelementptr double, ptr [[IN]], i64 32
; PTR64-NEXT:    [[COL_LOAD3:%.*]] = load <3 x double>, ptr [[VEC_GEP2]], align 8
; PTR64-NEXT:    [[TMP0:%.*]] = shufflevector <3 x double> [[COL_LOAD]], <3 x double> [[COL_LOAD1]], <6 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5>
; PTR64-NEXT:    [[TMP1:%.*]] = shufflevector <3 x double> [[COL_LOAD3]], <3 x double> poison, <6 x i32> <i32 0, i32 1, i32 2, i32 poison, i32 poison, i32 poison>
; PTR64-NEXT:    [[TMP2:%.*]] = shufflevector <6 x double> [[TMP0]], <6 x double> [[TMP1]], <9 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8>
; PTR64-NEXT:    ret <9 x double> [[TMP2]]
;
; PTR32-LABEL: @strided_load_3x3_const_stride_i128(
; PTR32-NEXT:  entry:
; PTR32-NEXT:    [[COL_LOAD:%.*]] = load <3 x double>, ptr [[IN:%.*]], align 8
; PTR32-NEXT:    [[VEC_GEP:%.*]] = getelementptr double, ptr [[IN]], i32 16
; PTR32-NEXT:    [[COL_LOAD1:%.*]] = load <3 x double>, ptr [[VEC_GEP]], align 8
; PTR32-NEXT:    [[VEC_GEP2:%.*]] = getelementptr double, ptr [[IN]], i32 32
; PTR32-NEXT:    [[COL_LOAD3:%.*]] = load <3 x double>, ptr [[VEC_GEP2]], align 8
; PTR32-NEXT:    [[TMP0:%.*]] = shufflevector <3 x double> [[COL_LOAD]], <3 x double> [[COL_LOAD1]], <6 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5>
; PTR32-NEXT:    [[TMP1:%.*]] = shufflevector <3 x double> [[COL_LOAD3]], <3 x double> poison, <6 x i32> <i32 0, i32 1, i32 2, i32 poison, i32 poison, i32 poison>
; PTR32-NEXT:    [[TMP2:%.*]] = shufflevector <6 x double> [[TMP0]], <6 x double> [[TMP1]], <9 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8>
; PTR32-NEXT:    ret <9 x double> [[TMP2]]
;
entry:
  %load = call <9 x double> @llvm.matrix.column.major.load.v9f64.i128(ptr %in, i128 16, i1 false, i32 3, i32 3)
  ret <9 x double> %load
}

define <9 x double> @strided_load_3x3_i64(ptr %in, i64 %stride) {
; PTR128-LABEL: @strided_load_3x3_i64(
; PTR128-NEXT:  entry:
; PTR128-NEXT:    [[STRIDE_CAST:%.*]] = zext i64 [[STRIDE:%.*]] to i128
; PTR128-NEXT:    [[VEC_START:%.*]] = mul i128 0, [[STRIDE_CAST]]
; PTR128-NEXT:    [[VEC_GEP:%.*]] = getelementptr double, ptr [[IN:%.*]], i128 [[VEC_START]]
; PTR128-NEXT:    [[COL_LOAD:%.*]] = load <3 x double>, ptr [[VEC_GEP]], align 8
; PTR128-NEXT:    [[VEC_START1:%.*]] = mul i128 1, [[STRIDE_CAST]]
; PTR128-NEXT:    [[VEC_GEP2:%.*]] = getelementptr double, ptr [[IN]], i128 [[VEC_START1]]
; PTR128-NEXT:    [[COL_LOAD3:%.*]] = load <3 x double>, ptr [[VEC_GEP2]], align 8
; PTR128-NEXT:    [[VEC_START4:%.*]] = mul i128 2, [[STRIDE_CAST]]
; PTR128-NEXT:    [[VEC_GEP5:%.*]] = getelementptr double, ptr [[IN]], i128 [[VEC_START4]]
; PTR128-NEXT:    [[COL_LOAD6:%.*]] = load <3 x double>, ptr [[VEC_GEP5]], align 8
; PTR128-NEXT:    [[TMP0:%.*]] = shufflevector <3 x double> [[COL_LOAD]], <3 x double> [[COL_LOAD3]], <6 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5>
; PTR128-NEXT:    [[TMP1:%.*]] = shufflevector <3 x double> [[COL_LOAD6]], <3 x double> poison, <6 x i32> <i32 0, i32 1, i32 2, i32 poison, i32 poison, i32 poison>
; PTR128-NEXT:    [[TMP2:%.*]] = shufflevector <6 x double> [[TMP0]], <6 x double> [[TMP1]], <9 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8>
; PTR128-NEXT:    ret <9 x double> [[TMP2]]
;
; PTR64-LABEL: @strided_load_3x3_i64(
; PTR64-NEXT:  entry:
; PTR64-NEXT:    [[VEC_START:%.*]] = mul i64 0, [[STRIDE:%.*]]
; PTR64-NEXT:    [[VEC_GEP:%.*]] = getelementptr double, ptr [[IN:%.*]], i64 [[VEC_START]]
; PTR64-NEXT:    [[COL_LOAD:%.*]] = load <3 x double>, ptr [[VEC_GEP]], align 8
; PTR64-NEXT:    [[VEC_START1:%.*]] = mul i64 1, [[STRIDE]]
; PTR64-NEXT:    [[VEC_GEP2:%.*]] = getelementptr double, ptr [[IN]], i64 [[VEC_START1]]
; PTR64-NEXT:    [[COL_LOAD3:%.*]] = load <3 x double>, ptr [[VEC_GEP2]], align 8
; PTR64-NEXT:    [[VEC_START4:%.*]] = mul i64 2, [[STRIDE]]
; PTR64-NEXT:    [[VEC_GEP5:%.*]] = getelementptr double, ptr [[IN]], i64 [[VEC_START4]]
; PTR64-NEXT:    [[COL_LOAD6:%.*]] = load <3 x double>, ptr [[VEC_GEP5]], align 8
; PTR64-NEXT:    [[TMP0:%.*]] = shufflevector <3 x double> [[COL_LOAD]], <3 x double> [[COL_LOAD3]], <6 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5>
; PTR64-NEXT:    [[TMP1:%.*]] = shufflevector <3 x double> [[COL_LOAD6]], <3 x double> poison, <6 x i32> <i32 0, i32 1, i32 2, i32 poison, i32 poison, i32 poison>
; PTR64-NEXT:    [[TMP2:%.*]] = shufflevector <6 x double> [[TMP0]], <6 x double> [[TMP1]], <9 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8>
; PTR64-NEXT:    ret <9 x double> [[TMP2]]
;
; PTR32-LABEL: @strided_load_3x3_i64(
; PTR32-NEXT:  entry:
; PTR32-NEXT:    [[STRIDE_CAST:%.*]] = trunc i64 [[STRIDE:%.*]] to i32
; PTR32-NEXT:    [[VEC_START:%.*]] = mul i32 0, [[STRIDE_CAST]]
; PTR32-NEXT:    [[VEC_GEP:%.*]] = getelementptr double, ptr [[IN:%.*]], i32 [[VEC_START]]
; PTR32-NEXT:    [[COL_LOAD:%.*]] = load <3 x double>, ptr [[VEC_GEP]], align 8
; PTR32-NEXT:    [[VEC_START1:%.*]] = mul i32 1, [[STRIDE_CAST]]
; PTR32-NEXT:    [[VEC_GEP2:%.*]] = getelementptr double, ptr [[IN]], i32 [[VEC_START1]]
; PTR32-NEXT:    [[COL_LOAD3:%.*]] = load <3 x double>, ptr [[VEC_GEP2]], align 8
; PTR32-NEXT:    [[VEC_START4:%.*]] = mul i32 2, [[STRIDE_CAST]]
; PTR32-NEXT:    [[VEC_GEP5:%.*]] = getelementptr double, ptr [[IN]], i32 [[VEC_START4]]
; PTR32-NEXT:    [[COL_LOAD6:%.*]] = load <3 x double>, ptr [[VEC_GEP5]], align 8
; PTR32-NEXT:    [[TMP0:%.*]] = shufflevector <3 x double> [[COL_LOAD]], <3 x double> [[COL_LOAD3]], <6 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5>
; PTR32-NEXT:    [[TMP1:%.*]] = shufflevector <3 x double> [[COL_LOAD6]], <3 x double> poison, <6 x i32> <i32 0, i32 1, i32 2, i32 poison, i32 poison, i32 poison>
; PTR32-NEXT:    [[TMP2:%.*]] = shufflevector <6 x double> [[TMP0]], <6 x double> [[TMP1]], <9 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8>
; PTR32-NEXT:    ret <9 x double> [[TMP2]]
;
entry:
  %load = call <9 x double> @llvm.matrix.column.major.load.v9f64.i64(ptr %in, i64 %stride, i1 false, i32 3, i32 3)
  ret <9 x double> %load
}

define <9 x double> @strided_load_3x3_const_stride_i64(ptr %in) {
; PTR128-LABEL: @strided_load_3x3_const_stride_i64(
; PTR128-NEXT:  entry:
; PTR128-NEXT:    [[COL_LOAD:%.*]] = load <3 x double>, ptr [[IN:%.*]], align 8
; PTR128-NEXT:    [[VEC_GEP:%.*]] = getelementptr double, ptr [[IN]], i128 16
; PTR128-NEXT:    [[COL_LOAD1:%.*]] = load <3 x double>, ptr [[VEC_GEP]], align 8
; PTR128-NEXT:    [[VEC_GEP2:%.*]] = getelementptr double, ptr [[IN]], i128 32
; PTR128-NEXT:    [[COL_LOAD3:%.*]] = load <3 x double>, ptr [[VEC_GEP2]], align 8
; PTR128-NEXT:    [[TMP0:%.*]] = shufflevector <3 x double> [[COL_LOAD]], <3 x double> [[COL_LOAD1]], <6 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5>
; PTR128-NEXT:    [[TMP1:%.*]] = shufflevector <3 x double> [[COL_LOAD3]], <3 x double> poison, <6 x i32> <i32 0, i32 1, i32 2, i32 poison, i32 poison, i32 poison>
; PTR128-NEXT:    [[TMP2:%.*]] = shufflevector <6 x double> [[TMP0]], <6 x double> [[TMP1]], <9 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8>
; PTR128-NEXT:    ret <9 x double> [[TMP2]]
;
; PTR64-LABEL: @strided_load_3x3_const_stride_i64(
; PTR64-NEXT:  entry:
; PTR64-NEXT:    [[COL_LOAD:%.*]] = load <3 x double>, ptr [[IN:%.*]], align 8
; PTR64-NEXT:    [[VEC_GEP:%.*]] = getelementptr double, ptr [[IN]], i64 16
; PTR64-NEXT:    [[COL_LOAD1:%.*]] = load <3 x double>, ptr [[VEC_GEP]], align 8
; PTR64-NEXT:    [[VEC_GEP2:%.*]] = getelementptr double, ptr [[IN]], i64 32
; PTR64-NEXT:    [[COL_LOAD3:%.*]] = load <3 x double>, ptr [[VEC_GEP2]], align 8
; PTR64-NEXT:    [[TMP0:%.*]] = shufflevector <3 x double> [[COL_LOAD]], <3 x double> [[COL_LOAD1]], <6 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5>
; PTR64-NEXT:    [[TMP1:%.*]] = shufflevector <3 x double> [[COL_LOAD3]], <3 x double> poison, <6 x i32> <i32 0, i32 1, i32 2, i32 poison, i32 poison, i32 poison>
; PTR64-NEXT:    [[TMP2:%.*]] = shufflevector <6 x double> [[TMP0]], <6 x double> [[TMP1]], <9 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8>
; PTR64-NEXT:    ret <9 x double> [[TMP2]]
;
; PTR32-LABEL: @strided_load_3x3_const_stride_i64(
; PTR32-NEXT:  entry:
; PTR32-NEXT:    [[COL_LOAD:%.*]] = load <3 x double>, ptr [[IN:%.*]], align 8
; PTR32-NEXT:    [[VEC_GEP:%.*]] = getelementptr double, ptr [[IN]], i32 16
; PTR32-NEXT:    [[COL_LOAD1:%.*]] = load <3 x double>, ptr [[VEC_GEP]], align 8
; PTR32-NEXT:    [[VEC_GEP2:%.*]] = getelementptr double, ptr [[IN]], i32 32
; PTR32-NEXT:    [[COL_LOAD3:%.*]] = load <3 x double>, ptr [[VEC_GEP2]], align 8
; PTR32-NEXT:    [[TMP0:%.*]] = shufflevector <3 x double> [[COL_LOAD]], <3 x double> [[COL_LOAD1]], <6 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5>
; PTR32-NEXT:    [[TMP1:%.*]] = shufflevector <3 x double> [[COL_LOAD3]], <3 x double> poison, <6 x i32> <i32 0, i32 1, i32 2, i32 poison, i32 poison, i32 poison>
; PTR32-NEXT:    [[TMP2:%.*]] = shufflevector <6 x double> [[TMP0]], <6 x double> [[TMP1]], <9 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8>
; PTR32-NEXT:    ret <9 x double> [[TMP2]]
;
entry:
  %load = call <9 x double> @llvm.matrix.column.major.load.v9f64.i64(ptr %in, i64 16, i1 false, i32 3, i32 3)
  ret <9 x double> %load
}

define <9 x double> @strided_load_3x3_i32(ptr %in, i32 %stride) {
; PTR128-LABEL: @strided_load_3x3_i32(
; PTR128-NEXT:  entry:
; PTR128-NEXT:    [[STRIDE_CAST:%.*]] = zext i32 [[STRIDE:%.*]] to i128
; PTR128-NEXT:    [[VEC_START:%.*]] = mul i128 0, [[STRIDE_CAST]]
; PTR128-NEXT:    [[VEC_GEP:%.*]] = getelementptr double, ptr [[IN:%.*]], i128 [[VEC_START]]
; PTR128-NEXT:    [[COL_LOAD:%.*]] = load <3 x double>, ptr [[VEC_GEP]], align 8
; PTR128-NEXT:    [[VEC_START1:%.*]] = mul i128 1, [[STRIDE_CAST]]
; PTR128-NEXT:    [[VEC_GEP2:%.*]] = getelementptr double, ptr [[IN]], i128 [[VEC_START1]]
; PTR128-NEXT:    [[COL_LOAD3:%.*]] = load <3 x double>, ptr [[VEC_GEP2]], align 8
; PTR128-NEXT:    [[VEC_START4:%.*]] = mul i128 2, [[STRIDE_CAST]]
; PTR128-NEXT:    [[VEC_GEP5:%.*]] = getelementptr double, ptr [[IN]], i128 [[VEC_START4]]
; PTR128-NEXT:    [[COL_LOAD6:%.*]] = load <3 x double>, ptr [[VEC_GEP5]], align 8
; PTR128-NEXT:    [[TMP0:%.*]] = shufflevector <3 x double> [[COL_LOAD]], <3 x double> [[COL_LOAD3]], <6 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5>
; PTR128-NEXT:    [[TMP1:%.*]] = shufflevector <3 x double> [[COL_LOAD6]], <3 x double> poison, <6 x i32> <i32 0, i32 1, i32 2, i32 poison, i32 poison, i32 poison>
; PTR128-NEXT:    [[TMP2:%.*]] = shufflevector <6 x double> [[TMP0]], <6 x double> [[TMP1]], <9 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8>
; PTR128-NEXT:    ret <9 x double> [[TMP2]]
;
; PTR64-LABEL: @strided_load_3x3_i32(
; PTR64-NEXT:  entry:
; PTR64-NEXT:    [[STRIDE_CAST:%.*]] = zext i32 [[STRIDE:%.*]] to i64
; PTR64-NEXT:    [[VEC_START:%.*]] = mul i64 0, [[STRIDE_CAST]]
; PTR64-NEXT:    [[VEC_GEP:%.*]] = getelementptr double, ptr [[IN:%.*]], i64 [[VEC_START]]
; PTR64-NEXT:    [[COL_LOAD:%.*]] = load <3 x double>, ptr [[VEC_GEP]], align 8
; PTR64-NEXT:    [[VEC_START1:%.*]] = mul i64 1, [[STRIDE_CAST]]
; PTR64-NEXT:    [[VEC_GEP2:%.*]] = getelementptr double, ptr [[IN]], i64 [[VEC_START1]]
; PTR64-NEXT:    [[COL_LOAD3:%.*]] = load <3 x double>, ptr [[VEC_GEP2]], align 8
; PTR64-NEXT:    [[VEC_START4:%.*]] = mul i64 2, [[STRIDE_CAST]]
; PTR64-NEXT:    [[VEC_GEP5:%.*]] = getelementptr double, ptr [[IN]], i64 [[VEC_START4]]
; PTR64-NEXT:    [[COL_LOAD6:%.*]] = load <3 x double>, ptr [[VEC_GEP5]], align 8
; PTR64-NEXT:    [[TMP0:%.*]] = shufflevector <3 x double> [[COL_LOAD]], <3 x double> [[COL_LOAD3]], <6 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5>
; PTR64-NEXT:    [[TMP1:%.*]] = shufflevector <3 x double> [[COL_LOAD6]], <3 x double> poison, <6 x i32> <i32 0, i32 1, i32 2, i32 poison, i32 poison, i32 poison>
; PTR64-NEXT:    [[TMP2:%.*]] = shufflevector <6 x double> [[TMP0]], <6 x double> [[TMP1]], <9 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8>
; PTR64-NEXT:    ret <9 x double> [[TMP2]]
;
; PTR32-LABEL: @strided_load_3x3_i32(
; PTR32-NEXT:  entry:
; PTR32-NEXT:    [[VEC_START:%.*]] = mul i32 0, [[STRIDE:%.*]]
; PTR32-NEXT:    [[VEC_GEP:%.*]] = getelementptr double, ptr [[IN:%.*]], i32 [[VEC_START]]
; PTR32-NEXT:    [[COL_LOAD:%.*]] = load <3 x double>, ptr [[VEC_GEP]], align 8
; PTR32-NEXT:    [[VEC_START1:%.*]] = mul i32 1, [[STRIDE]]
; PTR32-NEXT:    [[VEC_GEP2:%.*]] = getelementptr double, ptr [[IN]], i32 [[VEC_START1]]
; PTR32-NEXT:    [[COL_LOAD3:%.*]] = load <3 x double>, ptr [[VEC_GEP2]], align 8
; PTR32-NEXT:    [[VEC_START4:%.*]] = mul i32 2, [[STRIDE]]
; PTR32-NEXT:    [[VEC_GEP5:%.*]] = getelementptr double, ptr [[IN]], i32 [[VEC_START4]]
; PTR32-NEXT:    [[COL_LOAD6:%.*]] = load <3 x double>, ptr [[VEC_GEP5]], align 8
; PTR32-NEXT:    [[TMP0:%.*]] = shufflevector <3 x double> [[COL_LOAD]], <3 x double> [[COL_LOAD3]], <6 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5>
; PTR32-NEXT:    [[TMP1:%.*]] = shufflevector <3 x double> [[COL_LOAD6]], <3 x double> poison, <6 x i32> <i32 0, i32 1, i32 2, i32 poison, i32 poison, i32 poison>
; PTR32-NEXT:    [[TMP2:%.*]] = shufflevector <6 x double> [[TMP0]], <6 x double> [[TMP1]], <9 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8>
; PTR32-NEXT:    ret <9 x double> [[TMP2]]
;
entry:
  %load = call <9 x double> @llvm.matrix.column.major.load.v9f64.i32(ptr %in, i32 %stride, i1 false, i32 3, i32 3)
  ret <9 x double> %load
}

define <9 x double> @strided_load_3x3_const_stride_i32(ptr %in) {
; PTR128-LABEL: @strided_load_3x3_const_stride_i32(
; PTR128-NEXT:  entry:
; PTR128-NEXT:    [[COL_LOAD:%.*]] = load <3 x double>, ptr [[IN:%.*]], align 8
; PTR128-NEXT:    [[VEC_GEP:%.*]] = getelementptr double, ptr [[IN]], i128 16
; PTR128-NEXT:    [[COL_LOAD1:%.*]] = load <3 x double>, ptr [[VEC_GEP]], align 8
; PTR128-NEXT:    [[VEC_GEP2:%.*]] = getelementptr double, ptr [[IN]], i128 32
; PTR128-NEXT:    [[COL_LOAD3:%.*]] = load <3 x double>, ptr [[VEC_GEP2]], align 8
; PTR128-NEXT:    [[TMP0:%.*]] = shufflevector <3 x double> [[COL_LOAD]], <3 x double> [[COL_LOAD1]], <6 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5>
; PTR128-NEXT:    [[TMP1:%.*]] = shufflevector <3 x double> [[COL_LOAD3]], <3 x double> poison, <6 x i32> <i32 0, i32 1, i32 2, i32 poison, i32 poison, i32 poison>
; PTR128-NEXT:    [[TMP2:%.*]] = shufflevector <6 x double> [[TMP0]], <6 x double> [[TMP1]], <9 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8>
; PTR128-NEXT:    ret <9 x double> [[TMP2]]
;
; PTR64-LABEL: @strided_load_3x3_const_stride_i32(
; PTR64-NEXT:  entry:
; PTR64-NEXT:    [[COL_LOAD:%.*]] = load <3 x double>, ptr [[IN:%.*]], align 8
; PTR64-NEXT:    [[VEC_GEP:%.*]] = getelementptr double, ptr [[IN]], i64 16
; PTR64-NEXT:    [[COL_LOAD1:%.*]] = load <3 x double>, ptr [[VEC_GEP]], align 8
; PTR64-NEXT:    [[VEC_GEP2:%.*]] = getelementptr double, ptr [[IN]], i64 32
; PTR64-NEXT:    [[COL_LOAD3:%.*]] = load <3 x double>, ptr [[VEC_GEP2]], align 8
; PTR64-NEXT:    [[TMP0:%.*]] = shufflevector <3 x double> [[COL_LOAD]], <3 x double> [[COL_LOAD1]], <6 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5>
; PTR64-NEXT:    [[TMP1:%.*]] = shufflevector <3 x double> [[COL_LOAD3]], <3 x double> poison, <6 x i32> <i32 0, i32 1, i32 2, i32 poison, i32 poison, i32 poison>
; PTR64-NEXT:    [[TMP2:%.*]] = shufflevector <6 x double> [[TMP0]], <6 x double> [[TMP1]], <9 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8>
; PTR64-NEXT:    ret <9 x double> [[TMP2]]
;
; PTR32-LABEL: @strided_load_3x3_const_stride_i32(
; PTR32-NEXT:  entry:
; PTR32-NEXT:    [[COL_LOAD:%.*]] = load <3 x double>, ptr [[IN:%.*]], align 8
; PTR32-NEXT:    [[VEC_GEP:%.*]] = getelementptr double, ptr [[IN]], i32 16
; PTR32-NEXT:    [[COL_LOAD1:%.*]] = load <3 x double>, ptr [[VEC_GEP]], align 8
; PTR32-NEXT:    [[VEC_GEP2:%.*]] = getelementptr double, ptr [[IN]], i32 32
; PTR32-NEXT:    [[COL_LOAD3:%.*]] = load <3 x double>, ptr [[VEC_GEP2]], align 8
; PTR32-NEXT:    [[TMP0:%.*]] = shufflevector <3 x double> [[COL_LOAD]], <3 x double> [[COL_LOAD1]], <6 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5>
; PTR32-NEXT:    [[TMP1:%.*]] = shufflevector <3 x double> [[COL_LOAD3]], <3 x double> poison, <6 x i32> <i32 0, i32 1, i32 2, i32 poison, i32 poison, i32 poison>
; PTR32-NEXT:    [[TMP2:%.*]] = shufflevector <6 x double> [[TMP0]], <6 x double> [[TMP1]], <9 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8>
; PTR32-NEXT:    ret <9 x double> [[TMP2]]
;
entry:
  %load = call <9 x double> @llvm.matrix.column.major.load.v9f64.i32(ptr %in, i32 16, i1 false, i32 3, i32 3)
  ret <9 x double> %load
}

declare <9 x double> @llvm.matrix.column.major.load.v9f64.i128(ptr, i128, i1, i32, i32)
declare <9 x double> @llvm.matrix.column.major.load.v9f64.i64(ptr, i64, i1, i32, i32)
declare <9 x double> @llvm.matrix.column.major.load.v9f64.i32(ptr, i32, i1, i32, i32)