[llvm-branch-commits] [llvm] AMDGPU: Remove wavefrontsize64 feature from dummy target (PR #117410)
Matt Arsenault via llvm-branch-commits
llvm-branch-commits at lists.llvm.org
Sat Nov 23 08:06:11 PST 2024
https://github.com/arsenm updated https://github.com/llvm/llvm-project/pull/117410
>From 08b936f75b9e86b2b4aced3054fe072b3f534668 Mon Sep 17 00:00:00 2001
From: Matt Arsenault <Matthew.Arsenault at amd.com>
Date: Fri, 22 Nov 2024 17:21:18 -0800
Subject: [PATCH 1/2] AMDGPU: Remove wavefrontsize64 feature from dummy target
This is a refinement for the existing hack. With this,
the default target will have neither wavefrontsize feature
present, unless it was explicitly specified. That is,
getWavefrontSize() == 64 no longer implies +wavefrontsize64.
getWavefrontSize() == 32 does imply +wavefrontsize32.
Continue to assume the value is 64 with no wavesize feature.
This maintains the codegenable property without any code
that directly cares about the wavesize needs to worry about it.
Introduce an isWaveSizeKnown helper to check if we know the
wavesize is accurate based on having one of the features explicitly
set, or a known target-cpu.
I'm not sure what's going on in wave_any.s. It's testing what
happens when both wavesizes are enabled, but this is treated
as an error in codegen. We now treat wave32 as the winning
case, so some cases that were previously printed as vcc are now
vcc_lo.
---
llvm/lib/Target/AMDGPU/GCNProcessors.td | 4 +-
llvm/lib/Target/AMDGPU/GCNSubtarget.cpp | 16 +++----
llvm/lib/Target/AMDGPU/GCNSubtarget.h | 8 ++++
.../AMDGPU/MCTargetDesc/AMDGPUInstPrinter.cpp | 6 +--
llvm/test/MC/AMDGPU/wave_any.s | 48 +++++++++----------
5 files changed, 44 insertions(+), 38 deletions(-)
diff --git a/llvm/lib/Target/AMDGPU/GCNProcessors.td b/llvm/lib/Target/AMDGPU/GCNProcessors.td
index 508f2dd83108d9..a86c76bb6075e8 100644
--- a/llvm/lib/Target/AMDGPU/GCNProcessors.td
+++ b/llvm/lib/Target/AMDGPU/GCNProcessors.td
@@ -9,11 +9,11 @@
// The code produced for "generic" is only useful for tests and cannot
// reasonably be expected to execute on any particular target.
def : ProcessorModel<"generic", NoSchedModel,
- [FeatureWavefrontSize64, FeatureGDS, FeatureGWS]
+ [FeatureGDS, FeatureGWS]
>;
def : ProcessorModel<"generic-hsa", NoSchedModel,
- [FeatureWavefrontSize64, FeatureGDS, FeatureGWS, FeatureFlatAddressSpace]
+ [FeatureGDS, FeatureGWS, FeatureFlatAddressSpace]
>;
//===------------------------------------------------------------===//
diff --git a/llvm/lib/Target/AMDGPU/GCNSubtarget.cpp b/llvm/lib/Target/AMDGPU/GCNSubtarget.cpp
index 6233ca2eb4f1dd..51361b75940560 100644
--- a/llvm/lib/Target/AMDGPU/GCNSubtarget.cpp
+++ b/llvm/lib/Target/AMDGPU/GCNSubtarget.cpp
@@ -100,14 +100,16 @@ GCNSubtarget &GCNSubtarget::initializeSubtargetDependencies(const Triple &TT,
if (Gen == AMDGPUSubtarget::INVALID) {
Gen = TT.getOS() == Triple::AMDHSA ? AMDGPUSubtarget::SEA_ISLANDS
: AMDGPUSubtarget::SOUTHERN_ISLANDS;
- }
-
- if (!hasFeature(AMDGPU::FeatureWavefrontSize32) &&
- !hasFeature(AMDGPU::FeatureWavefrontSize64)) {
+ // Assume wave64 for the unknown target, if not explicitly set.
+ if (getWavefrontSizeLog2() == 0)
+ WavefrontSizeLog2 = 6;
+ } else if (!hasFeature(AMDGPU::FeatureWavefrontSize32) &&
+ !hasFeature(AMDGPU::FeatureWavefrontSize64)) {
// If there is no default wave size it must be a generation before gfx10,
// these have FeatureWavefrontSize64 in their definition already. For gfx10+
// set wave32 as a default.
ToggleFeature(AMDGPU::FeatureWavefrontSize32);
+ WavefrontSizeLog2 = getGeneration() >= AMDGPUSubtarget::GFX10 ? 5 : 6;
}
// We don't support FP64 for EG/NI atm.
@@ -147,10 +149,6 @@ GCNSubtarget &GCNSubtarget::initializeSubtargetDependencies(const Triple &TT,
!getFeatureBits().test(AMDGPU::FeatureCuMode))
LocalMemorySize *= 2;
- // Don't crash on invalid devices.
- if (WavefrontSizeLog2 == 0)
- WavefrontSizeLog2 = 5;
-
HasFminFmaxLegacy = getGeneration() < AMDGPUSubtarget::VOLCANIC_ISLANDS;
HasSMulHi = getGeneration() >= AMDGPUSubtarget::GFX9;
@@ -166,7 +164,7 @@ GCNSubtarget &GCNSubtarget::initializeSubtargetDependencies(const Triple &TT,
void GCNSubtarget::checkSubtargetFeatures(const Function &F) const {
LLVMContext &Ctx = F.getContext();
- if (hasFeature(AMDGPU::FeatureWavefrontSize32) ==
+ if (hasFeature(AMDGPU::FeatureWavefrontSize32) &&
hasFeature(AMDGPU::FeatureWavefrontSize64)) {
Ctx.diagnose(DiagnosticInfoUnsupported(
F, "must specify exactly one of wavefrontsize32 and wavefrontsize64"));
diff --git a/llvm/lib/Target/AMDGPU/GCNSubtarget.h b/llvm/lib/Target/AMDGPU/GCNSubtarget.h
index c7022ce78650ea..18219174b16b1e 100644
--- a/llvm/lib/Target/AMDGPU/GCNSubtarget.h
+++ b/llvm/lib/Target/AMDGPU/GCNSubtarget.h
@@ -1569,6 +1569,14 @@ class GCNSubtarget final : public AMDGPUGenSubtargetInfo,
return getWavefrontSize() == 64;
}
+ /// Returns if the wavesize of this subtarget is known reliable. This is false
+ /// only for the a default target-cpu that does not have an explicit
+ /// +wavefrontsize target feature.
+ bool isWaveSizeKnown() const {
+ return hasFeature(AMDGPU::FeatureWavefrontSize32) ||
+ hasFeature(AMDGPU::FeatureWavefrontSize64);
+ }
+
const TargetRegisterClass *getBoolRC() const {
return getRegisterInfo()->getBoolRC();
}
diff --git a/llvm/lib/Target/AMDGPU/MCTargetDesc/AMDGPUInstPrinter.cpp b/llvm/lib/Target/AMDGPU/MCTargetDesc/AMDGPUInstPrinter.cpp
index 344028c4b48689..e21aa70c9859a0 100644
--- a/llvm/lib/Target/AMDGPU/MCTargetDesc/AMDGPUInstPrinter.cpp
+++ b/llvm/lib/Target/AMDGPU/MCTargetDesc/AMDGPUInstPrinter.cpp
@@ -649,9 +649,9 @@ void AMDGPUInstPrinter::printDefaultVccOperand(bool FirstOperand,
raw_ostream &O) {
if (!FirstOperand)
O << ", ";
- printRegOperand(STI.hasFeature(AMDGPU::FeatureWavefrontSize64)
- ? AMDGPU::VCC
- : AMDGPU::VCC_LO,
+ printRegOperand(STI.hasFeature(AMDGPU::FeatureWavefrontSize32)
+ ? AMDGPU::VCC_LO
+ : AMDGPU::VCC,
O, MRI);
if (FirstOperand)
O << ", ";
diff --git a/llvm/test/MC/AMDGPU/wave_any.s b/llvm/test/MC/AMDGPU/wave_any.s
index 825a0abc172240..27502eff89bfc8 100644
--- a/llvm/test/MC/AMDGPU/wave_any.s
+++ b/llvm/test/MC/AMDGPU/wave_any.s
@@ -1,13 +1,13 @@
// RUN: llvm-mc -triple=amdgcn -mcpu=gfx1010 -mattr=+wavefrontsize32,+wavefrontsize64 -show-encoding %s | FileCheck --check-prefix=GFX10 %s
v_cmp_ge_i32_e32 s0, v0
-// GFX10: v_cmp_ge_i32_e32 vcc, s0, v0 ; encoding: [0x00,0x00,0x0c,0x7d]
+// GFX10: v_cmp_ge_i32_e32 vcc_lo, s0, v0 ; encoding: [0x00,0x00,0x0c,0x7d]
v_cmp_ge_i32_e32 vcc_lo, s0, v1
-// GFX10: v_cmp_ge_i32_e32 vcc, s0, v1 ; encoding: [0x00,0x02,0x0c,0x7d]
+// GFX10: v_cmp_ge_i32_e32 vcc_lo, s0, v1 ; encoding: [0x00,0x02,0x0c,0x7d]
v_cmp_ge_i32_e32 vcc, s0, v2
-// GFX10: v_cmp_ge_i32_e32 vcc, s0, v2 ; encoding: [0x00,0x04,0x0c,0x7d]
+// GFX10: v_cmp_ge_i32_e32 vcc_lo, s0, v2 ; encoding: [0x00,0x04,0x0c,0x7d]
v_cmp_le_f16_sdwa s0, v3, v4 src0_sel:WORD_1 src1_sel:DWORD
// GFX10: v_cmp_le_f16_sdwa s0, v3, v4 src0_sel:WORD_1 src1_sel:DWORD ; encoding: [0xf9,0x08,0x96,0x7d,0x03,0x80,0x05,0x06]
@@ -16,10 +16,10 @@ v_cmp_le_f16_sdwa s[0:1], v3, v4 src0_sel:WORD_1 src1_sel:DWORD
// GFX10: v_cmp_le_f16_sdwa s[0:1], v3, v4 src0_sel:WORD_1 src1_sel:DWORD ; encoding: [0xf9,0x08,0x96,0x7d,0x03,0x80,0x05,0x06]
v_cmp_class_f32_e32 vcc_lo, s0, v0
-// GFX10: v_cmp_class_f32_e32 vcc, s0, v0 ; encoding: [0x00,0x00,0x10,0x7d]
+// GFX10: v_cmp_class_f32_e32 vcc_lo, s0, v0 ; encoding: [0x00,0x00,0x10,0x7d]
v_cmp_class_f32_e32 vcc, s0, v0
-// GFX10: v_cmp_class_f32_e32 vcc, s0, v0 ; encoding: [0x00,0x00,0x10,0x7d]
+// GFX10: v_cmp_class_f32_e32 vcc_lo, s0, v0 ; encoding: [0x00,0x00,0x10,0x7d]
v_cmp_class_f16_sdwa vcc_lo, v1, v2 src0_sel:DWORD src1_sel:DWORD
// GFX10: v_cmp_class_f16_sdwa vcc_lo, v1, v2 src0_sel:DWORD src1_sel:DWORD ; encoding: [0xf9,0x04,0x1e,0x7d,0x01,0x00,0x06,0x06]
@@ -34,40 +34,40 @@ v_cmp_class_f16_sdwa s[0:1], v1, v2 src0_sel:DWORD src1_sel:DWORD
// GFX10: v_cmp_class_f16_sdwa s[0:1], v1, v2 src0_sel:DWORD src1_sel:DWORD ; encoding: [0xf9,0x04,0x1e,0x7d,0x01,0x80,0x06,0x06]
v_cndmask_b32_e32 v1, v2, v3,
-// GFX10: v_cndmask_b32_e32 v1, v2, v3, vcc ; encoding: [0x02,0x07,0x02,0x02]
+// GFX10: v_cndmask_b32_e32 v1, v2, v3, vcc_lo ; encoding: [0x02,0x07,0x02,0x02]
v_cndmask_b32_e32 v1, v2, v3, vcc_lo
-// GFX10: v_cndmask_b32_e32 v1, v2, v3, vcc ; encoding: [0x02,0x07,0x02,0x02]
+// GFX10: v_cndmask_b32_e32 v1, v2, v3, vcc_lo ; encoding: [0x02,0x07,0x02,0x02]
v_cndmask_b32_e32 v1, v2, v3, vcc
-// GFX10: v_cndmask_b32_e32 v1, v2, v3, vcc ; encoding: [0x02,0x07,0x02,0x02]
+// GFX10: v_cndmask_b32_e32 v1, v2, v3, vcc_lo ; encoding: [0x02,0x07,0x02,0x02]
v_add_co_ci_u32_e32 v3, vcc_lo, v3, v4, vcc_lo
-// GFX10: v_add_co_ci_u32_e32 v3, vcc, v3, v4, vcc ; encoding: [0x03,0x09,0x06,0x50]
+// GFX10: v_add_co_ci_u32_e32 v3, vcc_lo, v3, v4, vcc_lo ; encoding: [0x03,0x09,0x06,0x50]
v_add_co_ci_u32_e32 v3, vcc, v3, v4, vcc
-// GFX10: v_add_co_ci_u32_e32 v3, vcc, v3, v4, vcc ; encoding: [0x03,0x09,0x06,0x50]
+// GFX10: v_add_co_ci_u32_e32 v3, vcc_lo, v3, v4, vcc_lo ; encoding: [0x03,0x09,0x06,0x50]
v_add_co_ci_u32_e32 v3, v3, v4
-// GFX10: v_add_co_ci_u32_e32 v3, vcc, v3, v4, vcc ; encoding: [0x03,0x09,0x06,0x50]
+// GFX10: v_add_co_ci_u32_e32 v3, vcc_lo, v3, v4, vcc_lo ; encoding: [0x03,0x09,0x06,0x50]
v_sub_co_ci_u32_e32 v3, vcc_lo, v3, v4, vcc_lo
-// GFX10: v_sub_co_ci_u32_e32 v3, vcc, v3, v4, vcc ; encoding: [0x03,0x09,0x06,0x52]
+// GFX10: v_sub_co_ci_u32_e32 v3, vcc_lo, v3, v4, vcc_lo ; encoding: [0x03,0x09,0x06,0x52]
v_sub_co_ci_u32_e32 v3, vcc, v3, v4, vcc
-// GFX10: v_sub_co_ci_u32_e32 v3, vcc, v3, v4, vcc ; encoding: [0x03,0x09,0x06,0x52]
+// GFX10: v_sub_co_ci_u32_e32 v3, vcc_lo, v3, v4, vcc_lo ; encoding: [0x03,0x09,0x06,0x52]
v_sub_co_ci_u32_e32 v3, v3, v4
-// GFX10: v_sub_co_ci_u32_e32 v3, vcc, v3, v4, vcc ; encoding: [0x03,0x09,0x06,0x52]
+// GFX10: v_sub_co_ci_u32_e32 v3, vcc_lo, v3, v4, vcc_lo ; encoding: [0x03,0x09,0x06,0x52]
v_subrev_co_ci_u32_e32 v1, vcc_lo, 0, v1, vcc_lo
-// GFX10: v_subrev_co_ci_u32_e32 v1, vcc, 0, v1, vcc ; encoding: [0x80,0x02,0x02,0x54]
+// GFX10: v_subrev_co_ci_u32_e32 v1, vcc_lo, 0, v1, vcc_lo ; encoding: [0x80,0x02,0x02,0x54]
v_subrev_co_ci_u32_e32 v1, vcc, 0, v1, vcc
-// GFX10: v_subrev_co_ci_u32_e32 v1, vcc, 0, v1, vcc ; encoding: [0x80,0x02,0x02,0x54]
+// GFX10: v_subrev_co_ci_u32_e32 v1, vcc_lo, 0, v1, vcc_lo ; encoding: [0x80,0x02,0x02,0x54]
v_subrev_co_ci_u32_e32 v1, 0, v1
-// GFX10: v_subrev_co_ci_u32_e32 v1, vcc, 0, v1, vcc ; encoding: [0x80,0x02,0x02,0x54]
+// GFX10: v_subrev_co_ci_u32_e32 v1, vcc_lo, 0, v1, vcc_lo ; encoding: [0x80,0x02,0x02,0x54]
v_add_co_ci_u32_sdwa v1, vcc_lo, v1, v4, vcc_lo dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
// GFX10: v_add_co_ci_u32_sdwa v1, vcc_lo, v1, v4, vcc_lo dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD ; encoding: [0xf9,0x08,0x02,0x50,0x01,0x06,0x00,0x06]
@@ -76,7 +76,7 @@ v_add_co_ci_u32_sdwa v1, vcc, v1, v4, vcc dst_sel:DWORD dst_unused:UNUSED_PAD sr
// GFX10: v_add_co_ci_u32_sdwa v1, vcc, v1, v4, vcc dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD ; encoding: [0xf9,0x08,0x02,0x50,0x01,0x06,0x00,0x06]
v_add_co_ci_u32_sdwa v1, v1, v4 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-// GFX10: v_add_co_ci_u32_sdwa v1, vcc, v1, v4, vcc dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD ; encoding: [0xf9,0x08,0x02,0x50,0x01,0x06,0x00,0x06]
+// GFX10: v_add_co_ci_u32_sdwa v1, vcc_lo, v1, v4, vcc_lo dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD ; encoding: [0xf9,0x08,0x02,0x50,0x01,0x06,0x00,0x06]
v_sub_co_ci_u32_sdwa v1, vcc_lo, v1, v4, vcc_lo dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
// GFX10: v_sub_co_ci_u32_sdwa v1, vcc_lo, v1, v4, vcc_lo dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD ; encoding: [0xf9,0x08,0x02,0x52,0x01,0x06,0x00,0x06]
@@ -85,7 +85,7 @@ v_sub_co_ci_u32_sdwa v1, vcc, v1, v4, vcc dst_sel:DWORD dst_unused:UNUSED_PAD sr
// GFX10: v_sub_co_ci_u32_sdwa v1, vcc, v1, v4, vcc dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD ; encoding: [0xf9,0x08,0x02,0x52,0x01,0x06,0x00,0x06]
v_sub_co_ci_u32_sdwa v1, v1, v4 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-// GFX10: v_sub_co_ci_u32_sdwa v1, vcc, v1, v4, vcc dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD ; encoding: [0xf9,0x08,0x02,0x52,0x01,0x06,0x00,0x06]
+// GFX10: v_sub_co_ci_u32_sdwa v1, vcc_lo, v1, v4, vcc_lo dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD ; encoding: [0xf9,0x08,0x02,0x52,0x01,0x06,0x00,0x06]
v_subrev_co_ci_u32_sdwa v1, vcc_lo, v1, v4, vcc_lo dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
// GFX10: v_subrev_co_ci_u32_sdwa v1, vcc_lo, v1, v4, vcc_lo dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD ; encoding: [0xf9,0x08,0x02,0x54,0x01,0x06,0x00,0x06]
@@ -94,10 +94,10 @@ v_subrev_co_ci_u32_sdwa v1, vcc, v1, v4, vcc dst_sel:DWORD dst_unused:UNUSED_PAD
// GFX10: v_subrev_co_ci_u32_sdwa v1, vcc, v1, v4, vcc dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD ; encoding: [0xf9,0x08,0x02,0x54,0x01,0x06,0x00,0x06]
v_subrev_co_ci_u32_sdwa v1, v1, v4 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-// GFX10: v_subrev_co_ci_u32_sdwa v1, vcc, v1, v4, vcc dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD ; encoding: [0xf9,0x08,0x02,0x54,0x01,0x06,0x00,0x06]
+// GFX10: v_subrev_co_ci_u32_sdwa v1, vcc_lo, v1, v4, vcc_lo dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD ; encoding: [0xf9,0x08,0x02,0x54,0x01,0x06,0x00,0x06]
v_add_co_ci_u32 v1, sext(v1), sext(v4) dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-// GFX10: v_add_co_ci_u32_sdwa v1, vcc, sext(v1), sext(v4), vcc dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD ; encoding: [0xf9,0x08,0x02,0x50,0x01,0x06,0x08,0x0e]
+// GFX10: v_add_co_ci_u32_sdwa v1, vcc_lo, sext(v1), sext(v4), vcc_lo dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD ; encoding: [0xf9,0x08,0x02,0x50,0x01,0x06,0x08,0x0e]
v_add_co_ci_u32_sdwa v1, vcc_lo, sext(v1), sext(v4), vcc_lo dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
// GFX10: v_add_co_ci_u32_sdwa v1, vcc_lo, sext(v1), sext(v4), vcc_lo dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD ; encoding: [0xf9,0x08,0x02,0x50,0x01,0x06,0x08,0x0e]
@@ -106,7 +106,7 @@ v_add_co_ci_u32_sdwa v1, vcc, sext(v1), sext(v4), vcc dst_sel:DWORD dst_unused:U
// GFX10: v_add_co_ci_u32_sdwa v1, vcc, sext(v1), sext(v4), vcc dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD ; encoding: [0xf9,0x08,0x02,0x50,0x01,0x06,0x08,0x0e]
v_add_co_ci_u32_dpp v5, v1, v2 quad_perm:[0,1,2,3] row_mask:0x0 bank_mask:0x0
-// GFX10: v_add_co_ci_u32_dpp v5, vcc, v1, v2, vcc quad_perm:[0,1,2,3] row_mask:0x0 bank_mask:0x0 ; encoding: [0xfa,0x04,0x0a,0x50,0x01,0xe4,0x00,0x00]
+// GFX10: v_add_co_ci_u32_dpp v5, vcc_lo, v1, v2, vcc_lo quad_perm:[0,1,2,3] row_mask:0x0 bank_mask:0x0 ; encoding: [0xfa,0x04,0x0a,0x50,0x01,0xe4,0x00,0x00]
v_add_co_ci_u32_dpp v5, vcc_lo, v1, v2, vcc_lo quad_perm:[0,1,2,3] row_mask:0x0 bank_mask:0x0
// GFX10: v_add_co_ci_u32_dpp v5, vcc_lo, v1, v2, vcc_lo quad_perm:[0,1,2,3] row_mask:0x0 bank_mask:0x0 ; encoding: [0xfa,0x04,0x0a,0x50,0x01,0xe4,0x00,0x00]
@@ -189,8 +189,8 @@ v_subrev_co_ci_u32_e64 v4, s[0:1], v1, v5, s[2:3]
v_add_co_ci_u32_e64 v4, vcc_lo, v1, v5, s2
// GFX10: v_add_co_ci_u32_e64 v4, vcc_lo, v1, v5, s2 ; encoding: [0x04,0x6a,0x28,0xd5,0x01,0x0b,0x0a,0x00]
-v_add_co_ci_u32_e64 v4, vcc, v1, v5, s[2:3]
-// GFX10: v_add_co_ci_u32_e64 v4, vcc, v1, v5, s[2:3] ; encoding: [0x04,0x6a,0x28,0xd5,0x01,0x0b,0x0a,0x00]
+v_add_co_ci_u32_e64 v4, vcc_lo, v1, v5, s[2:3]
+// GFX10: v_add_co_ci_u32_e64 v4, vcc_lo, v1, v5, s[2:3] ; encoding: [0x04,0x6a,0x28,0xd5,0x01,0x0b,0x0a,0x00]
v_add_co_ci_u32_e64 v4, s0, v1, v5, vcc_lo
// GFX10: v_add_co_ci_u32_e64 v4, s0, v1, v5, vcc_lo ; encoding: [0x04,0x00,0x28,0xd5,0x01,0x0b,0xaa,0x01]
>From 79ff6e6f7e63108e90eb4737e8088331656305f3 Mon Sep 17 00:00:00 2001
From: Matt Arsenault <Matthew.Arsenault at amd.com>
Date: Fri, 22 Nov 2024 20:19:13 -0800
Subject: [PATCH 2/2] check if wave32 in disassembler
---
.../AMDGPU/Disassembler/AMDGPUDisassembler.cpp | 16 ++++++++--------
1 file changed, 8 insertions(+), 8 deletions(-)
diff --git a/llvm/lib/Target/AMDGPU/Disassembler/AMDGPUDisassembler.cpp b/llvm/lib/Target/AMDGPU/Disassembler/AMDGPUDisassembler.cpp
index 7817c5ff5acc0a..136fe2e3f90d02 100644
--- a/llvm/lib/Target/AMDGPU/Disassembler/AMDGPUDisassembler.cpp
+++ b/llvm/lib/Target/AMDGPU/Disassembler/AMDGPUDisassembler.cpp
@@ -1826,28 +1826,28 @@ MCOperand AMDGPUDisassembler::decodeSDWAVopcDst(unsigned Val) const {
STI.hasFeature(AMDGPU::FeatureGFX10)) &&
"SDWAVopcDst should be present only on GFX9+");
- bool IsWave64 = STI.hasFeature(AMDGPU::FeatureWavefrontSize64);
+ bool IsWave32 = STI.hasFeature(AMDGPU::FeatureWavefrontSize32);
if (Val & SDWA9EncValues::VOPC_DST_VCC_MASK) {
Val &= SDWA9EncValues::VOPC_DST_SGPR_MASK;
int TTmpIdx = getTTmpIdx(Val);
if (TTmpIdx >= 0) {
- auto TTmpClsId = getTtmpClassId(IsWave64 ? OPW64 : OPW32);
+ auto TTmpClsId = getTtmpClassId(IsWave32 ? OPW32 : OPW64);
return createSRegOperand(TTmpClsId, TTmpIdx);
}
if (Val > SGPR_MAX) {
- return IsWave64 ? decodeSpecialReg64(Val) : decodeSpecialReg32(Val);
+ return IsWave32 ? decodeSpecialReg32(Val) : decodeSpecialReg64(Val);
}
- return createSRegOperand(getSgprClassId(IsWave64 ? OPW64 : OPW32), Val);
+ return createSRegOperand(getSgprClassId(IsWave32 ? OPW32 : OPW64), Val);
}
- return createRegOperand(IsWave64 ? AMDGPU::VCC : AMDGPU::VCC_LO);
+ return createRegOperand(IsWave32 ? AMDGPU::VCC_LO : AMDGPU::VCC);
}
MCOperand AMDGPUDisassembler::decodeBoolReg(unsigned Val) const {
- return STI.hasFeature(AMDGPU::FeatureWavefrontSize64)
- ? decodeSrcOp(OPW64, Val)
- : decodeSrcOp(OPW32, Val);
+ return STI.hasFeature(AMDGPU::FeatureWavefrontSize32)
+ ? decodeSrcOp(OPW32, Val)
+ : decodeSrcOp(OPW64, Val);
}
MCOperand AMDGPUDisassembler::decodeSplitBarrier(unsigned Val) const {
More information about the llvm-branch-commits
mailing list