similar to: [LLVMdev] Stack alignment in kernel

On Jan 9, 2012, at 10:00 AM, Jakob Stoklund Olesen wrote: > > On Jan 8, 2012, at 11:18 PM, Demikhovsky, Elena wrote: > >> I'll explain what we see in the code. >> 1. The caller saves XMM registers across the call if needed (according to DEFS definition). >> YMMs are not in the set, so caller does not take care. > > This is not how the register allocator

Hi Elena, You're correct. LLVM does not align the stack to 32-bytes for AVX and unaligned moves should be used for YMM spills. I wrote some code to align the stack to 32-bytes when AVX spills are present; it does break the x86-64 ABI though. If upstream would be interested in this code, I can arrange with my employer to send a patch to the mailing list. -Cameron On Mar 1, 2012, at 4:09 PM,

This is the wrong code: declare <16 x float> @foo(<16 x float>) define <16 x float> @test(<16 x float> %x, <16 x float> %y) nounwind { entry: %x1 = fadd <16 x float> %x, %y %call = call <16 x float> @foo(<16 x float> %x1) nounwind %y1 = fsub <16 x float> %call, %y ret <16 x float> %y1 } ./llc -mattr=+avx

./llc -mattr=+avx -stack-alignment=16 < basic.ll | grep movaps | grep ymm | grep rbp vmovaps -176(%rbp), %ymm14 vmovaps -144(%rbp), %ymm11 vmovaps -240(%rbp), %ymm13 vmovaps -208(%rbp), %ymm9 vmovaps -272(%rbp), %ymm7 vmovaps -304(%rbp), %ymm0 vmovaps -112(%rbp), %ymm0 vmovaps -80(%rbp), %ymm1 vmovaps -112(%rbp), %ymm0

On Thu, Mar 1, 2012 at 5:30 PM, Cameron McInally <cameron.mcinally at nyu.edu>wrote: > Aligning the stack to 32 bytes when there are auto AVX vector variables >> present shouldn't necessarily break the x86-64 ABI, as long as smaller auto >> variables remain properly aligned. A similar approach was taken for i386 >> in GCC in order to support SSE vectors. >>

On Thu, Mar 1, 2012 at 4:29 PM, Evandro Menezes <emenezes at codeaurora.org>wrote: ... > Aligning the stack to 32 bytes when there are auto AVX vector variables > present shouldn't necessarily break the x86-64 ABI, as long as smaller auto > variables remain properly aligned. A similar approach was taken for i386 > in GCC in order to support SSE vectors. > > Perhaps

Cameron, Aligning the stack to 32 bytes when there are auto AVX vector variables present shouldn't necessarily break the x86-64 ABI, as long as smaller auto variables remain properly aligned. A similar approach was taken for i386 in GCC in order to support SSE vectors. Perhaps you could elaborate where the ABI was violated when your patch is applied. HTH -- Evandro Menezes

- we used the C4/C5 (first prefix) byte instead of the apparent ModR/M one as the second prefix byte - early decoding normalized vex.reg, thus corrupting it for the main consumer (copy_REX_VEX()), resulting in #UD on the two-operand instructions we emulate Also add respective test cases to the testing utility plus - fix get_fpu() (the fall-through order was inverted) - add cpu_has_avx2,

I wonder why AVX is not used automatically if available at the host machine. In contrast to that, SSE41 instructions (like pmulld) are automatically used if the host machine supports SSE41. E.g. $ cat avx.ll define void @_fun1(<8 x float>*, <8 x float>*) { _L1: %x = load <8 x float>* %0 %y = load <8 x float>* %1 %z = fadd <8 x float> %x, %y store

On Thu, 24 May 2012, Pan, Wei wrote: > Very likely AVX is not enabled in your llc. This feature was enabled > just recently (late of April). I forgot to mention that I am using recent LLVM-3.1 and in principle my llc knows about avx as I have shown in the second example. But avx does not seem to be used by default. On Thu, 24 May 2012, Henning Thielemann wrote: > $ llc -o - -mattr

Hi, Please consider the following loop: using v4f32 = float __attribute__((__vector_size__(16))); void fct6(v4f32 *x) { #pragma clang loop vectorize(enable) for (int i = 0; i < 256; ++i) x[i] = 7 * x[i]; } After compiling it with: clang++ -O3 -march=native -mtune=native \ -Rpass=loop-vectorize,slp-vectorize -Rpass-missed=loop-vectorize,slp-vectorize

On Thu, Mar 01, 2012 at 06:16:46PM +0000, Demikhovsky, Elena wrote: > vmovaps should not access stack if it is not aligned to 32 I'm not completely sure I understand your problem. Are you saying that the generated code assumes 256bit alignment, your default stack alignment is 128bit and LLVM doesn't adjust it automatically? Joerg

Hi, I am using mcjit in llvm 3.6 to jit kernels to x86 avx2. I've noticed some inefficient use of the stack around constant vectors. In one example, I have code that computes a series of constant vectors at compile time. Each vector has a single use. In the final asm, I see a series of spills at the top of the function of all the constant vectors immediately to stack, then each use references

When stack is unaligned, LLVM should generate vmovups instead of vmovaps. - Elena -----Original Message----- From: llvmdev-bounces at cs.uiuc.edu [mailto:llvmdev-bounces at cs.uiuc.edu] On Behalf Of Joerg Sonnenberger Sent: Thursday, March 01, 2012 20:31 To: llvmdev at cs.uiuc.edu Subject: Re: [LLVMdev] Stack alignment on X86 AVX seems incorrect On Thu, Mar 01, 2012 at 06:16:46PM +0000,

Illustrative Example: clang -fveclib=SVML -O3 svml.c -mavx #include <math.h> void foo(double *a, int N){ int i; #pragma clang loop vectorize_width(8) for (i=0;i<N;i++){ a[i] = sin(i); } } Currently, this results in a call to <8 x double> __svml_sin8(<8 x double>) after the vectorizer. This is 8-element SVML sin() called with 8-element argument. On the surface,

Even if you explicitly specify –stack-alignment=16 the aligned movs are still generated. It is not an issue related to ABI. See my original mail: ./llc -mattr=+avx -stack-alignment=16 < basic.ll | grep movaps | grep ymm | grep rbp vmovaps -176(%rbp), %ymm14 vmovaps -144(%rbp), %ymm11 vmovaps -240(%rbp), %ymm13 - Elena From: Cameron McInally

Hi Elena, Thank you very much for looking in to that. I'll go ahead and remove the isCommutable flag from those instructions, since it sounds like that's the right thing to do. I would still like to change the default from the 231 variant to 213 too, as this will reduce code-size for accumulator-style loops. I have at least one benchmark that shows significant speedups when this change

I'm seeing a difference in how LLVM 3.3 and 3.2 emit unaligned vector loads on AVX. 3.3 is splitting up an unaligned vector load but in 3.2, it was emitted as a single instruction (details below). In a matrix-matrix inner-kernel, I see a ~25% decrease in performance, which seems to be due to this. Any ideas why this changed? Thanks! Zach LLVM Code: define <4 x double> @vstore(<4 x

Thank you, Clayton. This is very helpful. We use the LLDB specific GDB remote extensions, and our debugger server supports "qRegisterInfo" package. "reg 0x3c" is the frame pointer. In the example mentioned above, we have SP = FP - 40 for current call frame. And variable "a" is stored at address (FP + -24) from asm instruction [FP + -24] = R3;; Thus we can conclude

Hi Kay, My patch will partially address your bug. For now I'm just looking to switch the default FMA from vfmadd213xx to vfmadd231xx. That will cause the code in PR17229 to compile as desired, but would regress code like: double foo(double a, double b, double c) { return a * b + c; } Which will now require a vmovaps + vfmadd231. If this impacts real benchmarks we could add an