From: Giacomo Travaglini Date: Tue, 26 Mar 2019 18:02:11 +0000 (+0000) Subject: arch-arm: Fix index generation for VecElem operands X-Git-Tag: v19.0.0.0~1004 X-Git-Url: https://git.libre-soc.org/?a=commitdiff_plain;h=93ad0d4324acd674f1877a1146c58412a03b4c39;p=gem5.git arch-arm: Fix index generation for VecElem operands Current operand generation is not providing VecElems with the right vector index and element index. The bug was covered when registers were 128 bit wide, but with SVE we have augmented the vector register size and the bug has been exposed. E.g. With dest = 2, FpDestP2 = (vec_index = 0, elem_index = 4) whereas it should be FpDestP2 = (vec_index = 1, elem_index = 0) Change-Id: Iad02fb477afd0d3dd3d437bf2ca4338fbd142107 Signed-off-by: Giacomo Travaglini Reviewed-by: Ciro Santilli Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/17710 --- diff --git a/src/arch/arm/isa/operands.isa b/src/arch/arm/isa/operands.isa index fb3e4de35..0f656dac9 100644 --- a/src/arch/arm/isa/operands.isa +++ b/src/arch/arm/isa/operands.isa @@ -266,14 +266,14 @@ def operands {{ #Abstracted floating point reg operands 'FpDest': vectorElem('dest / 4', 'dest % 4'), - 'FpDestP0': vectorElem('dest / 4', '(dest % 4) + 0'), - 'FpDestP1': vectorElem('dest / 4', '(dest % 4) + 1'), - 'FpDestP2': vectorElem('dest / 4', '(dest % 4) + 2'), - 'FpDestP3': vectorElem('dest / 4', '(dest % 4) + 3'), - 'FpDestP4': vectorElem('(dest / 4) + 1', '(dest % 4) + 0'), - 'FpDestP5': vectorElem('(dest / 4) + 1', '(dest % 4) + 1'), - 'FpDestP6': vectorElem('(dest / 4) + 1', '(dest % 4) + 2'), - 'FpDestP7': vectorElem('(dest / 4) + 1', '(dest % 4) + 3'), + 'FpDestP0': vectorElem('(dest + 0) / 4', '(dest + 0) % 4'), + 'FpDestP1': vectorElem('(dest + 1) / 4', '(dest + 1) % 4'), + 'FpDestP2': vectorElem('(dest + 2) / 4', '(dest + 2) % 4'), + 'FpDestP3': vectorElem('(dest + 3) / 4', '(dest + 3) % 4'), + 'FpDestP4': vectorElem('(dest + 4) / 4', '(dest + 4) % 4'), + 'FpDestP5': vectorElem('(dest + 5) / 4', '(dest + 5) % 4'), + 'FpDestP6': vectorElem('(dest + 6) / 4', '(dest + 6) % 4'), + 'FpDestP7': vectorElem('(dest + 7) / 4', '(dest + 7) % 4'), 'FpDestS0P0': vectorElem( '(dest + step * 0 + 0) / 4', '(dest + step * 0 + 0) % 4'), @@ -293,20 +293,20 @@ def operands {{ '(dest + step * 3 + 1) / 4', '(dest + step * 3 + 1) % 4'), 'FpDest2': vectorElem('dest2 / 4', 'dest2 % 4'), - 'FpDest2P0': vectorElem('dest2 / 4', '(dest2 % 4) + 0'), - 'FpDest2P1': vectorElem('dest2 / 4', '(dest2 % 4) + 1'), - 'FpDest2P2': vectorElem('dest2 / 4', '(dest2 % 4) + 2'), - 'FpDest2P3': vectorElem('dest2 / 4', '(dest2 % 4) + 3'), + 'FpDest2P0': vectorElem('(dest2 + 0) / 4', '(dest2 + 0) % 4'), + 'FpDest2P1': vectorElem('(dest2 + 1) / 4', '(dest2 + 1) % 4'), + 'FpDest2P2': vectorElem('(dest2 + 2) / 4', '(dest2 + 2) % 4'), + 'FpDest2P3': vectorElem('(dest2 + 3) / 4', '(dest2 + 3) % 4'), 'FpOp1': vectorElem('op1 / 4', 'op1 % 4'), - 'FpOp1P0': vectorElem('op1 / 4', '(op1 % 4) + 0'), - 'FpOp1P1': vectorElem('op1 / 4', '(op1 % 4) + 1'), - 'FpOp1P2': vectorElem('op1 / 4', '(op1 % 4) + 2'), - 'FpOp1P3': vectorElem('op1 / 4', '(op1 % 4) + 3'), - 'FpOp1P4': vectorElem('(op1 / 4) + 1', '(op1 % 4) + 0'), - 'FpOp1P5': vectorElem('(op1 / 4) + 1', '(op1 % 4) + 1'), - 'FpOp1P6': vectorElem('(op1 / 4) + 1', '(op1 % 4) + 2'), - 'FpOp1P7': vectorElem('(op1 / 4) + 1', '(op1 % 4) + 3'), + 'FpOp1P0': vectorElem('(op1 + 0) / 4', '(op1 + 0) % 4'), + 'FpOp1P1': vectorElem('(op1 + 1) / 4', '(op1 + 1) % 4'), + 'FpOp1P2': vectorElem('(op1 + 2) / 4', '(op1 + 2) % 4'), + 'FpOp1P3': vectorElem('(op1 + 3) / 4', '(op1 + 3) % 4'), + 'FpOp1P4': vectorElem('(op1 + 4) / 4', '(op1 + 4) % 4'), + 'FpOp1P5': vectorElem('(op1 + 5) / 4', '(op1 + 5) % 4'), + 'FpOp1P6': vectorElem('(op1 + 6) / 4', '(op1 + 6) % 4'), + 'FpOp1P7': vectorElem('(op1 + 7) / 4', '(op1 + 7) % 4'), 'FpOp1S0P0': vectorElem( '(op1 + step * 0 + 0) / 4', '(op1 + step * 0 + 0) % 4'), @@ -326,10 +326,10 @@ def operands {{ '(op1 + step * 3 + 1) / 4', '(op1 + step * 3 + 1) % 4'), 'FpOp2': vectorElem('op2 / 4', 'op2 % 4'), - 'FpOp2P0': vectorElem('op2 / 4', '(op2 % 4) + 0'), - 'FpOp2P1': vectorElem('op2 / 4', '(op2 % 4) + 1'), - 'FpOp2P2': vectorElem('op2 / 4', '(op2 % 4) + 2'), - 'FpOp2P3': vectorElem('op2 / 4', '(op2 % 4) + 3'), + 'FpOp2P0': vectorElem('(op2 + 0) / 4', '(op2 + 0) % 4'), + 'FpOp2P1': vectorElem('(op2 + 1) / 4', '(op2 + 1) % 4'), + 'FpOp2P2': vectorElem('(op2 + 2) / 4', '(op2 + 2) % 4'), + 'FpOp2P3': vectorElem('(op2 + 3) / 4', '(op2 + 3) % 4'), # Create AArch64 unpacked view of the FP registers # Name ::= 'AA64Vec' OpSpec [LaneSpec]